Molecular Cell, Vol. 4, 915–924, December, 1999, Copyright 1999 by Cell Press

Selective Requirement for Src Kinasesduring VEGF-Induced Angiogenesisand Vascular Permeability

forms of Src or intact C-terminal Src kinase (Csk) todisrupt endogenous Src activity within the chick chorio-allantoic membrane (CAM) or mouse skin to directlyevaluate the general role of Src kinases during angiogen-esis. Evidence is provided here that Src kinase is re-

Brian P. Eliceiri,* Robert Paul,*Pamela L. Schwartzberg,† John D. Hood,*Jie Leng,* and David A. Cheresh*‡

*Departments of Immunology and Vascular BiologyThe Scripps Research Institute10550 North Torrey Pines Road quired for VEGF-, but not bFGF-, mediated angiogenesis

in both the chick embryo and the mouse. In fact, SrcLa Jolla, California 92037†The National Human Genome Research Institute kinase activity was found to be required for endothelial

cell survival during VEGF-mediated angiogenesis. WhileNational Institutes of HealthBethesda, Maryland 20892 VEGF is an endothelial cell mitogen (Ferrara and Davis-

Smyth, 1997), it was originally described for its vascularpermeability (VP) activity (Senger et al., 1983; Connollyet al., 1989). In fact, VEGF is unique in this regard, asSummaryother growth factors such as bFGF can induce neovas-cularization but do not induce vascular permeabilitySrc kinase activity was found to protect endothelial

cells from apoptosis during vascular endothelial growth (Connolly et al., 1989; Murohara et al., 1998). An analysisof mice deficient in specific SFKs demonstrates no de-factor (VEGF)–, but not basic fibroblast growth factor

(bFGF)–, mediated angiogenesis in chick embryos and crease in VEGF-dependent neovascularization but acomplete ablation of its VP activity in src2/2 or yes2/2mice. In fact, retroviral targeting of kinase-deleted Src

to tumor-associated blood vessels suppressed angio- mice, while fyn2/2 mice show no such defect. While micelacking Src show no VP response to VEGF, they do showgenesis and the growth of a VEGF-producing tumor.

Although mice lacking individual Src family kinases a VP response to an inflammatory mediator. Therefore,multiple SFKs can serve as key signaling intermediates(SFKs) showed normal angiogenesis, mice deficient in

pp60c-src or pp62c-yes showed no VEGF-induced vascular involved in VEGF-induced vascular proliferation, whilethe VP activity of this growth factor depends on Src orpermeability (VP), yet fyn2/2 mice displayed normal

VP. In contrast, inflammation-mediated VP appeared Yes in particular.normal in Src-deficient mice. Therefore, VEGF-, butnot bFGF-, mediated angiogenesis requires SFK activ- Resultsity in general, whereas the VP activity of VEGF specifi-cally depends on the SFKs, Src, or Yes. Src Activity Is Required for VEGF-, but Not bFGF-,

Induced AngiogenesisTo establish whether endogenous Src activity was asso-

Introduction ciated with growth factor–mediated angiogenesis, filterdisks saturated with either bFGF or VEGF were placed

SFKs are important signaling molecules that respond to on the CAM of 10-day-old chick embryos. This treatmenta wide range of stimuli including growth factors (Twam- is known to promote a robust angiogenic response asley-Stein et al., 1993; Broome and Hunter, 1996) and measured after 48–72 hr (Brooks et al., 1994a). Lysatesadhesion proteins in the extracellular matrix (Kaplan et of these CAMs were evaluated for Src activity by immu-al., 1994; Schwartz et al., 1995; Thomas and Brugge, noprecipitating Src and measuring its ability to phos-1997; Klinghoffer et al., 1999). Once activated, SFKs phorylate a GST–focal adhesion kinase (FAK) fusion pro-affect a wide range of downstream signaling events in- tein in an in vitro kinase assay. At least a 2-fold increasecluding the activation of MAP kinases (Courtneidge et in endogenous Src activity was detected in these lysatesal., 1993). While in vitro studies have elucidated a role 120 min after the addition of either bFGF or VEGF tofor Src in cellular function, due to mechanisms of redun- the CAM tissue (Figure 1A). Importantly, we observed adancies and compensation, mice lacking a single SFK similar increase in Src activity 15 min after the addition(Soriano et al., 1991; Stein et al., 1994) have provided of either growth factor (data not shown). To determinelimited insight into the biological function of this impor- whether Src activity was required for angiogenesis,tant family of nonreceptor tyrosine kinases. CAMs stimulated with either bFGF or VEGF were in-

Previous studies have implicated SFKs in vascular fected with an avian-specific retroviral vector (RCAS)cell proliferation. For example, v-Src, an oncogenic vari- containing a truncation mutant of Src lacking its C-termi-ant of Src, is known to promote hemangioma formation nal kinase domain (Src 251) (Kaplan et al., 1994). Thisin chicks (Stoker et al., 1990), suggesting that under Src 251 and similar truncation mutants have been shownnormal circumstances, c-Src or other SFKs may regulate to function as a dominant negative of multiple SFKs,the growth of blood vessels. To initially address this thereby blocking signaling events downstream of growthissue, we used avian- or murine-targeted retroviral deliv- factor receptors (Broome and Hunter, 1996; P. L. S.,ery systems to express mutationally active or inactive unpublished data). The RCAS retrovirus, when applied

to CAM tissues, infects fibroblasts and endothelial cellsproximal to the filter disk as determined by infecting‡ To whom correspondence should be addressed (e-mail: cheresh@

scripps.edu). CAMs with an RCAS–GFP vector and examination by

Molecular Cell916

Figure 1. Activation of Endogenous Src Ki-nase Activity by bFGF and VEGF and the Ef-fect of Kinase-Deleted Src on AngiogenesisIn Vivo

(A) Tissue extracts of 10-day-old chick CAMswere exposed to filter paper disks saturatedwith bFGF or VEGF (2 mg/ml) for 2 hr. Endoge-nous Src was immunoprecipitated from equiv-alent amounts of total protein and subjectedto an in vitro immune complex kinase assayswith a FAK–GST fusion protein as a substrate,electrophoresed, and transferred to nitrocel-lulose. The relative fold increase in Src activ-ity is indicated in italics. The above kinaseassay blot was probed with an anti-Src anti-body as a loading control for equivalent Srcand IgG content.(B) Chick CAMs (9 day) were exposed to filterpaper disks saturated with RCAS–Src 251 (ki-nase deleted) or RCAS–GFP containing re-troviruses or buffer for 20 hr and then incu-bated in the presence or absence of bFGF orVEGF for an additional 72 hr. Tissue extractsof these CAMs were examined for endoge-nous Src activity by in vitro immune complexkinase assay as described above using FAK–GST as a substrate.(C) The level of angiogenesis was quantifiedin embryos incubated with RCAS–Src251 orRCAS–GFP followed by stimulation with ei-ther bFGF or VEGF as described above.Blood vessels were enumerated by countingblood vessel branch points in a doubleblinded manner. Each bar represents themean 6 SEM of three replicates.(D) Micrographs of representative CAMs weretaken with an Olympus stereomicroscope.Scale bar, 350 mm.

confocal microscopy (data not shown). Delivery of this reports that VEGF and bFGF stimulate distinct pathwaysof angiogenesis (Friedlander et al., 1995; Ziche et al.,kinase-deleted Src completely disrupted endogenous

Src kinase activity in these tissues induced with either 1997).growth factor (Figure 1B).

To examine the role of Src in angiogenesis, CAMs Suppression of Human Tumor Growth by Targetingthe Tumor Vascular Compartment withstimulated with either bFGF or VEGF were infected with

the Src 251–containing retrovirus. As shown in Figure Retroviral Delivery of Src 251Tumor growth depends on angiogenesis (Weidner et al.,1C, angiogenesis, as measured 72 hr after stimulation

with VEGF, was suppressed by delivery of Src 251; how- 1991; Folkman and Shing, 1992; Brooks et al., 1994b).In fact, recent reports suggest that tumor growth is sus-ever, to our surprise, bFGF-induced angiogenesis was

not affected. Importantly, equivalent levels of viral infec- ceptible to the antiangiogenic effects of VEGF receptorantagonists (Kim et al., 1993). Therefore, experimentstion were detected in VEGF- and bFGF-stimulated CAMs

as measured by epifluorescence and immunoblot analy- were designed to determine whether suppression of an-giogenesis by delivery of kinase-deleted Src 251 wouldsis of GFP and Src 251, respectively (data not shown).

The inhibition of VEGF-induced angiogenesis by kinase- influence the growth of a human medulloblastoma(DAOY), a highly angiogenic tumor known to producedeleted Src was likely due to a direct effect on endothe-

lial cells, since VEGF is a known endothelial cell-specific VEGF and very little bFGF (data not shown). This humantumor readily grows on the CAM and produces an activemitogen. In addition, the failure of Src 251 to disrupt

bFGF-induced angiogenesis indicates that the effects angiogenic response (Figure 2), allowing us to selec-tively target the tumor vasculature by using the avian-on VEGF-mediated angiogenesis are not due to general

toxicity. Together, these results demonstrate that, while specific RCAS retrovirus, without infecting the humanmedulloblastoma cells. Delivery of RCAS containing Srcboth bFGF and VEGF can activate Src kinase in these

tissues, only VEGF-induced blood vessel formation re- 251 to preestablished medulloblastomas resulted in aselective expression of the virus in the tumor-associatedquired this activity. These findings support the recent

Src Requirement for Angiogenesis and Permeability917

fewer in number compared to the tumor vessels incontrol animals (Figure 2C). The fact that RCAS-GFP-infected tumors showed GFP localization only in thetumor vasculature suggests that the antitumor effectsobserved with retrovirally delivered Src 251 were dueto its targeting and antiangiogenic properties.

Src Requirement for Endothelial Cell Survivalduring VEGF-, but Not bFGF-,Mediated AngiogenesisRecent evidence suggests that growth factor receptors(Choi and Ballermann, 1995; Satake et al., 1998) andintegrins (Meredith et al., 1993; Brooks et al., 1994a)promote survival of angiogenic endothelial cells. Thefact that both growth factors and adhesion receptorsalso regulate Src activity prompted us to examine therole of Src in endothelial cell survival during angiogen-esis. Furthermore, the Src 251 mutant has been foundto induce apoptosis in selective cell types during bonedevelopment (P. L. S., L. Xing, and B. Boyce, unpub-lished data). CAMs stimulated with either bFGF or VEGFwere infected with retrovirus containing Src 251, andcryostat sections of these tissues were examined forthe presence of apoptotic cells. As shown in Figure 3A,delivery of Src 251 promoted extensive TUNEL stainingamong the factor VIII–related antigen (von Willebrandfactor [vWf]) positive blood vessels in VEGF-, but notbFGF-, stimulated CAMS. In fact, minimal apoptosis wasobserved among other cell types in these CAMs (Figure3), suggesting an endothelial cell-specific requirementfor Src kinase activity for cell survival in VEGF-activatedblood vessels. In a second series of experiments, retro-virus-infected CAMs stimulated with VEGF or bFGF weresubjected to limited collagenase digestion to prepare asingle cell suspension. These CAM-derived cells wereshown to contain approximately 20%–50% endothelialcells (vWf positive) (Figures 3C and 3D) and analyzedfor apoptosis by flow cytometric detection of annexinFigure 2. Retroviral Delivery of RCAS–Src 251 to Human TumorsV–positive cells, an early apoptosis marker. As shownGrowing on the Chick CAM Reverses Tumor Growthin Figure 3B, cells derived from VEGF-stimulated CAMs(A) Human DAOY medulloblastomas, which express VEGF, were

grown on the CAM of chick embryos as described in the Experimen- infected with Src 251 had significantly increased an-tal Procedures. Retrovirus containing RCAS–GFP or RCAS–Src 251 nexin V staining relative to cells from mock RCAS-GFP-was topically applied to preestablished tumors of greater than 50 infected CAMs treated with VEGF. In contrast, cells de-mg. A representative micrograph of a medulloblastoma tumor frag- rived from mock-infected CAMs or those infected withment infected with RCAS–GFP expressing GFP reveals exclusive

RCAS–Src 251 and stimulated with bFGF exhibited littleexpression in the tumor blood vessels (arrowhead) as detected byor no annexin V staining (data not shown). In addition,optical sectioning with a Bio-Rad 1024 laser confocal scanning mi-

croscope. Scale bar, 500 mm. no annexin V staining was detected among cells derived(B) Tumors treated as above were allowed to grow for 3 or 6 days, from nonstimulated or bFGF-stimulated CAMs (data notafter which they were resected and wet weights were determined. shown). These data demonstrate that Src kinase activityData are expressed as the mean change in tumor weight (from the is selectively required for endothelial cell survival during50 mg tumor starting weight) 6 SEM of two replicates. RCAS–Src

VEGF, but not bFGF-mediated angiogenesis in the CAM.251 had a significant impact on tumor growth after 3 days (*p ,

0.002) and 6 days (**p , 0.05).(C) Representative stereomicrographs of medulloblastoma tumors Selective Requirement for Src Kinase Activitysurgically removed from the embryos were taken with an Olympus in a Subcutaneous Murine Modelstereomicroscope (scale bar, 350 mm). (Lower panel) A high magnifi- of Angiogenesiscation micrograph of each tumor showing the vasculature in detail

To further analyze the role of Src in angiogenesis, a(scale bar, 350 mm). The arrowhead indicates blood vessel disruptionmurine model was employed. In this case, angiogenesisin RCAS–Src 251–treated tumors.was induced by subcutaneous injection of growth fac-tor–depleted Matrigel supplemented with either bFGF(400 ng/ml) or VEGF (400 ng/ml) in athymic wehi (nu/nu)blood vessels (Figure 2A), which led to a complete sup-

pression of tumor growth (Figure 2B). Importantly, the adult mice and analyzed after 5 days (Passaniti et al.,1992). Angiogenesis was quantitated by removing andtumor-associated blood vessels in animals treated with

virus containing Src 251 were severely disrupted and extracting the angiogenic tissue and then subjecting the

Molecular Cell918

Figure 3. Apoptosis in VEGF-Stimulated Blood Vessels Expressing Src 251

(A) Immunolocalization of factor VIII–related antigen (von Willlebrand factor), apoptag immunostaining of apoptotic cells, and nuclear stainingwith DAPI in cryosections of CAMs expressing RCAS–Src 251 or RCAS–GFP, after stimulation with bFGF or VEGF as described in Figure 1.The merge represents an overlay of the factor VIII staining and apoptag staining. The fluorescence from the GFP was not preserved in thefixation protocol used for the indirect immunofluorescence in these experiments. These micrographs were representative of blood vesselstaining in duplicate samples. Scale bar, 50 mm.(B) Apoptotic cells were identified by annexin V staining of RCAS–Src 251–infected CAMS treated with VEGF and detected by flow cytometry.Collagenase-dissociated cells isolated from RCAS–Src 251– (black) or RCAS-GFP- (mock, white) infected CAMs treated with VEGF, asdescribed in Figure 1, were incubated with annexin V. The fluorescence from the GFP was not detected in these assays, and the FACS profilewas similar to untreated controls. The flow cytometry data for each experiment was representative of at least three replicates.(C) Anti-vWf staining was detected with a FITC-labeled secondary antibody used to identify endothelial cells by flow cytometry, and this wascompared to parallel collagenase-dissociated untreated CAM cells incubated without primary antibody.(D) Immunolocalization of endogenous von Willebrand factor in collagenase-dissociated untreated permeabilized CAM cells (arrowhead)replated on 3 mg/ml collagen and detected with a fluorescent secondary antibody (bar, 10 mm).

lysates to immunoblotting with a VEGF receptor anti- the number of positively stained CD34 blood vessels ineach cryosection (Figure 4C).body (flk-1) (Figure 4A) that is specific for endothelial

cells. As observed in the chick, expression of the kinase-deleted Src 251 cDNA blocked VEGF-induced angiogen-esis in this murine model while having no effect on bFGF- The Effect of Intradermal Expression of VEGF

in src2/2 or src1/2 Mice Earsinduced angiogenesis (Figure 4B). To establish the roleof endogenous Src in this model, tissues were infected To extend the observations made in the chicken and

mouse angiogenesis models, a direct genetic approachwith a retrovirus expressing Csk that inhibits endoge-nous Src activity by phosphorylation of the C-terminal was employed to examine intradermal VEGF-induced

angiogenesis in src2/2 mice. We also considered the factregulatory site (Nada et al., 1991). Expression of Cskblocked VEGF-, but not bFGF-, induced angiogenesis that VEGF both initiates new blood vessel growth and

can promote vascular permeability (Senger et al., 1983;(Figure 4), confirming a role for endogenous Src activityin VEGF-mediated angiogenesis. Neovascularization of Ferrara and Davis-Smyth, 1997). Intradermal injections

of adenovirus expressing a human VEGF cDNA werethese virus-infected VEGF-stimulated tissues was con-firmed by direct immunofluorescence with a FITC-conju- performed in the ear of src1/2 and src2/2, while control

b-galactosidase expressing adenovirus was injectedgated anti-CD34 antibody (Figure 4) or an anti-flk-1 anti-body (data not shown) and quantitated by enumerating into the opposite ear of each mouse. VEGF-dependent

Src Requirement for Angiogenesis and Permeability919

5A, which confirms the extent of the vascular leakagein src1/2 mice that is essentially absent in the src2/2

mice. The vascular leakage in these animals suggestedthat the VP activity, which has been associated withangiogenesis in vivo (Dvorak et al., 1995), could be selec-tively disrupted in pp60c-src-deficient mice.

VEGF Fails to Compromise the Blood–BrainBarrier in Mice Lacking pp60c-src

The brain vasculature is characterized by a highly re-strictive blood–brain barrier that prohibits small mole-cules from extravasating into the surrounding brain tis-sue. Tumor growth within the brain can compromisethis barrier due in part to the production of angiogenicgrowth factors such as VEGF. Therefore, we examinedthe nature of the blood–brain barrier in src1/2 or src2/2

mice. In this case, VEGF or saline was stereotacticallyinjected into the right or left hemisphere of the brain,respectively. All mice received systemic injections ofEvan’s blue to monitor VP activity. As shown in Figure5B, vascular leakage of blood was localized to the VEGF-injected hemisphere in src1/2 mice, but there a completeabsence of vascular leakage in src2/2 mice. This wasalso the case when examining the VP by measuringthe accumulation of Evan’s blue dye as detected byepifluorescence analysis of cryostat sections of thesebrains (Figure 5C). Thus, VEGF compromises the blood–brain barrier in a manner that depends on pp60c-src.

VEGF-Mediated VP, but Not Inflammation-Associated VP, Depends on pp60c-src

Figure 4. Retroviral Delivery of Src 251 and Csk in a SubcutaneousTo further analyze and quantitate the effect of VEGF asMurine Angiogenesis Modela VP factor in src1/2 or src2/2 mice, we used the Miles(A) Angiogenesis was induced by a subcutaneous injection of growthassay (Miles and Miles, 1952) to quantitatively measurefactor-depleted Matrigel containing saline or VEGF (400 ng/ml) with

2 3 106 ecotropic packaging cells expressing GFP retrovirus in the the vascular permeability in the skin of these animals.flank of athymic wehi (nu/nu) mice and analyzed after 5 days of incuba- VEGF was injected intradermally in src1/2 or src2/2 micetion. The neovascularization was quantitated by immunoblotting that had received an intravenous systemic administra-with a VEGF receptor antibody (flk-1) that is specific for endothelial

tion of Evan’s blue dye. Within 15 min after injection ofcells.VEGF, there was a 3-fold increase in VP in src1/2. How-(B) The effects of kinase-deleted Src 251, Csk, or GFP retrovirus onever, in src2/2 mice, we observed no detectable VP activ-VEGF- (400 ng/ml) or bFGF- (400 ng/ml) induced angiogenesis was

analyzed by immunoblotting the tissue lysates with an anti-flk-1 ity (Figures 6A and 6B). Dye elution of the injected skinantibody. patches was quantitated and compared with control(C) The effect of the Src 251– and Csk-expressing retroviruses on saline and bFGF (Figure 6B, left panel). bFGF or salineVEGF-induced neovascularization was quantified by enumerating

controls injected adjacent to the VEGF showed no signif-the number of CD34 positive vessels in tissue cross sections byicant increase in VP.indirect immunofluorescence in triplicate random fields at 203 as

Vascular leakage/permeability is also known to occurdescribed in the Experimental Procedures.during inflammation, which allows for the accumulationof serum-associated adhesive protein and inflammatorycells in tissues. In fact, inflammatory mediators them-new blood vessel growth in src1/2 ears was first detect-selves directly promote vascular leakage. Therefore, weable within 48 hr, and neovascularization was analyzedtested one such inflammatory mediator, allyl isothiocya-after 5 days (Figure 5A). There were identical viral ex-nate, also known as mustard oil (Inoue et al., 1997), inpression levels in src1/2 and src2/2 as determined bysrc1/2 or src2/2 mice for its capacity to produce VP.X-gal staining of b-galactosidase-adenovirus injectedUnlike that observed in VEGF-stimulated src2/2 animals,ears (data not shown). In VEGF-injected src2/2 ears, therewe detected no decrease in the VP produced by injectionwas no significant decrease in angiogenesis (data notof the inflammatory mediator allyl isothiocyanate (Figureshown) as measured by counting branch points (p ,6B, right panel). Thus, we conclude that Src plays a0.05). However, the most apparent phenotype in theseselective role in the VP activity induced with VEGF andanimals was the complete blockade in the vascular leak-does not influence VP associated with the inflammatoryage compared to the VEGF-injected src1/2 ears. Repre-

sentative ears injected with VEGF are shown in Figure process.

Molecular Cell920

Figure 5. The Effect of VEGF-Induced Vascu-lar Leakage in the Ears and Brains of src2/2

and src1/2 Mice

(A) Gene delivery of the human VEGF cDNAin an adenovirus vector was injected intrader-mally in the right ear of src1/2 or src2/2 mice,and the neovascularization of the ears werephotographed after 5 days of expression. Ad-enovirus expressing b-galactosidase was in-jected into the left ears as a negative control.Staining for b-galactosidase in these earsconfirmed similar adenovirus expression ineach genetic background. Scale bar, 1 mm;n 5 4.(B) VEGF or saline was stereotactically in-jected into the left or right frontal lobes, re-spectively, of src1/2 or src2/2. After injectionwith Evan’s blue and perfusion, the brainswere removed and photographed with a ste-reoscope (63, final magnification; arrow-head, injection site).(C) Cross sections of the above VEGF- orsaline-injected brains from src1/2 or src2/2

mice were prepared and analyzed for VEGF-induced VP by confocal microscopy to visual-ize the fluorescence of the extravasated Ev-an’s blue (63, final magnification; arrowhead,injection site).

VEGF-Mediated VP Activity Depends on Src naling events required for the growth of new blood ves-and Yes but Not Fyn sels. In this report, evidence is provided that two angio-We next tested the specificity of the Src requirement genic growth factors, bFGF and VEGF, initiate signalingfor VP by examining the VEGF-induced VP activity asso- pathways that can be distinguished based on their re-ciated with SFKs such as Fyn or Yes, which, like Src, quirement for Src kinase activity. Even though bothare known to be expressed in endothelial cells (Bull et bFGF and VEGF led to increased Src activity in angio-al., 1994; Kiefer et al., 1994). In fact, we confirmed that genic tissues, only VEGF-induced angiogenesis dependedthese three SFKs were expressed equivalently in the on it. This was based on studies where kinase-deletedaortas of wild-type mice (data not shown). Like src2/2 Src or Csk was retrovirally delivered to stimulated bloodmice, animals deficient in Yes were also defective in vessels. The use of intact Csk was important as it blocksVEGF-induced VP (Figure 6C). However, to our surprise, the activity of endogenous Src rather than acting as amice lacking Fyn retained a high VP in response to VEGF dominant-negative mutant like Src 251. Src activity wasthat was not significantly different from control animals found to be required for the survival of VEGF-stimulated(Figure 6C). The disruption of VEGF-induced vascular endothelial cells in vivo.permeability in src2/2 or yes2/2 mice demonstrates that VEGF was originally described as a vascular perme-the kinase activity of specific SFKs is essential for VEGF- ability factor secreted by tumor cells (Senger et al.,mediated signaling event leading to VP activity but not 1983). Using mice deficient for specific SFKs, we dem-angiogenesis. onstrated that pp60c-src or pp62c-yes are essential for

VEGF-induced VP, while its angiogenic activity was notsignificantly influenced in these animals. Moreover, ani-Discussionmals deficient in Fyn show no loss of VP activity demon-strating that only certain SFKs are required to regulateWhile multiple growth factors and adhesion events can

promote angiogenesis, little is known regarding the sig- VEGF-mediated VP activity. Importantly, all three of

Src Requirement for Angiogenesis and Permeability921

other postnatal process. Interestingly, mice lacking thecombination of Src, Yes, and Fyn or the VEGF receptorshow embryonic lethality by day 9.5, a time during devel-opment that is characterized by active vasculogenesis(Fong et al., 1995; Shalaby et al., 1995). This, togetherwith the fact that mice lacking individual SFKs developnormal appearing blood vessels, suggests that compen-sation can take place among these SFKs. This is sup-ported by our observation that suppression of Src kinaseactivity in general by Csk or Src251 suppressed neovas-cularization in mice or chick embryos in response toVEGF while individual SFK knockouts develop normally.

Evidence provided in this study demonstrates thatVEGF and bFGF potentiate somewhat different biologi-cal and biochemical effects during the early stages ofangiogenesis. There may be a physiological rationale forthe existence of two angiogenic pathways. For example,blood vessels in various organs may differ with respectto distinct ECM-associated adhesive proteins and/orgrowth factors. Neovascularization in the retina hasbeen linked to VEGF expression (D’Amore, 1994; Milleret al., 1994), while that induced during cutaneous woundrepair has been associated with bFGF (Takenaka et al.,1997). This may allow endothelial cells to meet the spe-cific needs of a given tissue depending on local require-ments for nutrients, oxygen, or waste elimination. Afterhypoxic injury, VEGF levels are known to rise immedi-ately (reviewed in Ferrara and Davis-Smyth, 1997). Per-haps this hypoxic response facilitates an immediate in-creased oxygenation by providing local vascular leakageprior to the actual formation of a new vascular network.This would predict that adult pp60c-src- or pp62c-yes-defi-Figure 6. Miles Assay for Vascular Permeability of VEGF in the Skincient mice may be less capable of restoring oxygenationof Mice Deficient in Src, Fyn, or Yesto damaged or hypoxic tissue. In fact, we noted that(A) The vascular permeability properties of VEGF in the skin of src1/2

stereotactic injection of VEGF in the brain could compro-(upper) or src2/2 (lower) mice was determined by intradermal injec-tion of saline or VEGF (400 ng) into mice that have been intravenously mise the blood–brain barrier in control animals. How-injected with Evan’s blue dye. After 15 min, skin patches were photo- ever, animals deficient in pp60c-src showed no breakdowngraphed (scale bar, 1 mm). Arrowheads indicate the injection sites. of the blood–brain barrier.(B) The regions surrounding the injection sites of the VEGF, bFGF,

VEGF is an angiogenic growth factor in many tumors.or saline were dissected, and the permeability quantitated by elutionIn fact, an anti-VEGF antibody (Kim et al., 1993) thatof the Evan’s blue in formamide at 568C for 24 hr, and the absorbanceblocks tumor growth in mice is being evaluated clinicallymeasured at 600 nm (left). The ability of an inflammation mediator

(allyl isothiocyanate), known to induce inflammation-related VP, was in patients with late-stage cancer. Given the strong as-tested in src1/2 or src2/2 mice (right). sociation between VEGF and tumor angiogenesis, our(C) The ability of VEGF to induce VP was compared in src2/2, fyn2/2, results may provide another approach to disrupt theor yes2/2 mice in the Miles assay. Data for each of the Miles assays

growth of tumors. Thus, by using an avian-specific retro-are expressed as the mean 6 SD of triplicate animals. src2/2 andvirus, we were able to specifically target the chick vascu-yes2/2 VP defects compared to control animals were statisticallylature of a growing human medulloblastoma. Evensignificant (*p , 0.05, paired t test), whereas the VP defects in

neither the VEGF-treated fyn2/2 mice nor the allyl isothiocyanate– though the tumor cells remained uninfected by the retro-treated src2/2 mice were statistically significant (**p , 0.05). virus, we observed suppressed tumor growth demon-

strating the potential therapeutic efficacy of this ap-proach.these SFKs were shown to be equivalently expressed

In a combination of experiments using retrovirally de-in the aortas, skin, and brain of wild-type mice and arelivered mutant Src and Csk as well as a direct analysesknown to be expressed in endothelial cells (Bull et al.,of src2/2 mice, we provide evidence that the Src tyrosine1994; Kiefer et al., 1994). To our surprise, inflammation-kinase family distinguishes two pathways of angiogen-induced VP was shown to be independent of Src kinaseesis. During VEGF-induced angiogenesis, SFK activityin these mice, suggesting that the VP activity inducedcontributes to endothelial cell survival. Furthermore, theduring inflammation and that induced upon VEGF stimu-VEGF-induced VP is dependent on SFKs, Src, or Yes,lation are regulated by distinct signaling pathways.but not Fyn, and the VP response is specific for VEGFMice lacking pp60c-src and pp62c-yes show apparentlyin contrast to inflammation-induced VP. Therefore, whilenormal vascular development, even though mice lackingSFKs serve compensatory roles during embryogenesisVEGF or its receptor die during development. Thus,and angiogenesis, VEGF-, but not bFGF-, mediated an-VEGF-induced VP activity is not required for develop-

ment. However, it may play a role in wound repair or giogenesis requires Src kinase activity for endothelial

Molecular Cell922

goat anti-mouse secondary antibodies as previously described (Eli-cell survival, whereas VP activity of VEGF depends onceiri et al., 1998).the SFKs, Src, or Yes.

In Vitro Kinase Assay for Src KinaseExperimental ProceduresThe kinase activity of endogenous Src kinase was assayed by theability of immunopurified Src to phosphorylate a FAK–GST fusionAntibodies and Reagentsprotein in an in vitro assay. Src was immunoprecipitated as de-A rabbit polyclonal antibody raised against amino acids 3–18 ofscribed above and subjected to a kinase assay, and the sampleshuman Src (N-16; Santa Cruz Biotechnology, Santa Cruz, CA) waswere analyzed by 15% SDS-PAGE and quantitated as describedused for immunoprecipitation for in vitro kinase assays, and mono-previously (Eliceiri et al., 1998).clonal antibody against avian pp60c-src (Upstate Biotechnology, Lake

Placid, NY) was used for Western blotting as a loading control forImmunostaining and Annexin V Labeling of Apoptotic Cellsthe kinase assays. The Src constructs were obtained from Dr. H.Cryosections of CAMs treated with RCAS–GFP or RCAS–Src 251Varmus (NIH), FAK–GST fusion protein was from Dr. D. Schlaepfertreated with bFGF or VEGF were analyzed for apoptotic cells using(The Scripps Research Institute [TSRI]), the DF-1 virus producer cellthe Apoptag kit (Oncor, Gaithersburg, MD). Sections were also im-line was from Dr. D. Foster (Univeristy of Minnesota), the DAOYmunostained with a rabbit polyclonal anti-vWF (Biogenix, San Ra-medulloblastoma cell line from Dr. W. Laug (Children’s Hospital,mon, CA) and counterstained with 1 mg/ml DAPI. Fluorescent imagesUSC, Los Angeles), RCAS–GFP was from Dr. C. Cepko (Harvard),were captured with a cooled CCD camera (Roper, Trenton, NJ),and bFGF was kindly provided by Dr. J. Abraham (Scios, Mountainand the fluorescent images were processed and exposure matchedView, CA). All other reagents and media were from Sigma-Aldrichbetween experimental treatments as previously described (Eliceiri(St Louis, MO) unless otherwise stated.et al., 1998).

To measure the apoptotic index of retrovirus-infected CAM tis-Src Constructs and Retrovirusessues, FITC-conjugated annexin V (Clontech, Palo Alto, CA) was usedFor the studies in the chick embryo, the replication competentto stain cell suspensions, and the washed cells were analyzed byRCASBP(A) (Hughes et al., 1987) retrovirus was used to expressflow cytometry. Cell suspensions of CAM cells were prepared frommutant Src cDNAs subcloned as NotI–ClaI. These constructs weremock- or virus-infected CAMs by digestion with 0.1% (w/v) collage-transfected into the chicken immortalized fibroblast line, DF-1. Viralnase type IV (Worthington Biochemicals, Lakewood, NJ) in RPMIsupernatants were collected from DF-1 producer cell lines in serum-1640 of minced CAM tissue rocking for 1 hr at 378C as previouslyfree CLM media. Viral supernatants were concentrated by ultracen-described (Brooks et al., 1994b) and filtered through 100 mM nylontrifugation at 48C for 2 hr at 22,000 rpm, and the pellets were resus-mesh (Becton Dickinson, Fountain Lakes, NJ). Fluorescence waspended in 1/100 the original volume in serum-free media with a titermeasured with a FACscan flow cytometer (Becton Dickinson) toof at least 108 i.u. (infectious units)/ml and stored at 2808C.count 10,000 cells.For the retrovirus studies in the subcutaneous murine matrigel

Measurement of vWf staining by FACS was performed with paral-angiogenesis assay, GFP, kinase-deleted Src 251, and Csk cDNAlel collagenase digested CAM tissue cell preparations, that werewas subcloned into the replication-defective murine Moloney retro-fixed in 1.6% paraformaldehyde, permeabilized in 70% ethanol, in-virus (pLNCX) vector. These constructs were transiently transfectedcubated the anti-vWf antibody, and detected with a FITC-conju-into the ecotropic producer line to generate cell-free titers of 105–106

gated secondary antibody.i.u./ml. Therefore, to increase the effective titer over the 5 day timecourse of the angiogenesis assay in the matrigel plug, the virus-

Tumor Growth Assaypackaging cells expressing the appropriate construct were includedThe 3 and 6 day DAOY medulloblastoma tumor growth assays werealong in the Matrigel to increase the retrovirus infection levels.performed in the chick CAM essentially as previously described(Brooks et al., 1994b). DAOY cells (5 3 106) were seeded on theVEGF AdenovirusCAM of a 10 day embryo. After 7 days, 50 mg tumor fragments wereRecombinant VEGF adenovirus was generated by cloning the hu-dissected and reseeded on another 10 day embryo and incubatedman VEGF cDNA from a human placenta cDNA library into pAd/CIfor another 3 or 6 days with the topical application (25 ml) of either(J. L., A. Reddy, and D. A. C., unpublished data) and cotransfectingcontrol RCAS–GFP retrovirus, RCAS–Src 251, or mock treatment.with pJM17 into an E1 transcomplementing 293 cell line as pre-Tumor resections and weighing were performed in a double blindviously described (Bett et al., 1994). High titer virus was isolated,manner removing only the easily definable solid tumor mass (Brookspurified, and titered to 1011 pfu/ml as previously described (Changet al., 1994b). The wet tumor weights after 3 or 6 days were com-et al., 1995). High titer clones were selected based on their expres-pared with initial weight, and the percent change of tumor weightsion of soluble VEGF secreted into the media of COS-7, endothelialwas determined for each group.cells, and in chick CAMs infected with the VEGF adenovirus (data

not shown).Immunofluorescence and MicroscopyCryosections of the plugs were also subjected to immunofluores-Chick Embryo Treatmentscent staining with an anti-CD34 antibody or an anti-flk antibody,Fertilized chick embryos (standard pathogen free grade; SPAFAS,photographed, and quantitated as described above for the CAMPreston, CT) were prepared, and the CAM was exposed as pre-angiogenesis assays.viously described (Eliceiri et al., 1998). For growth factor–only experi-

Whole-mount direct fluorescence of RCAS-GFP-infected tumorments, cortisone acetate–soaked filter disks were soaked with 250fragment was accomplished by dissecting a tumor fragment andng of bFGF or VEGF for 2 hr before harvest. For virus experimentsimaging the unfixed tissue directly on a slide with a laser confocalon the CAM, disks were soaked in 20 ml of viral stock per disk.microscope (MRC 1024; Bio-Rad, Hercules, CA).These disks were applied to the CAM of 9 day chick embryos and

incubated at 378C for 24 hr. Then, either serum-free media or growthfactors were added at a concentration of 5 mg/ml to the CAM in 20 Murine Matrigel Angiogenesis Assayml of the virus stock as an additional boost of virus to the CAM Growth factor–depleted Matrigel (Becton Dickinson) (400 ml) supple-tissue and incubated for an additional 72 hr. CAM assays was quanti- mented with PBS, bFGF (400 ng/ml), or VEGF (400 ng/ml) (Passanititated by counting branch points as described previously (Eliceiri et et al., 1992) and murine-specific ecotropic packaging cells (φNX-al., 1998) in triplicate samples in a double blind manner. Eco; G. Nolan, Stanford) producing retrovirus expressing GFP, Src

251, or Csk cDNAs was injected subcutaneously in 6-week-old maleathymic wehi (nu/nu) mice. The plugs remained palpable for 5 days,Immunoprecipitation and Immunoblotting

CAM tissues were homogenized in a RIPA lysis buffer, used for facilitating a direct resection of the plug for further analysis by immu-noblotting of plug homogenates or immunostaining of plug cryosec-immunoprecipitations or immunoblots as previously described (Eli-

ceiri et al., 1998). Anti-Src and anti-flk1 antibodies used in immu- tions. The accuracy of the quantitative methods was confirmed byspectrophotometric analysis of homogenates of plugs from animalsmoblots were detected with horseradish peroxidase-conjugated

Src Requirement for Angiogenesis and Permeability923

that had been intravenously injected with FITC-labeled lectin (J. D. H. Adenovirus-mediated over-expression of the cyclin/cyclin-depen-dent kinase inhibitor, p21 inhibits vascular smooth muscle cell prolif-and D. A. C., unpublished data).eration and neointima formation in the rat carotid artery model ofballoon angioplasty. J. Clin. Invest. 96, 2260–2268.Intradermal Ear Injections and Miles Assay

pp60c-src-, pp59c-fyn-, and pp62c-yes-deficient mice (129/Sv/Ev 3 Choi, M.E., and Ballermann, B.J. (1995). Inhibition of capillary mor-phogenesis and associated apoptosis by dominant negative mutantC57Bl6/J) were generated as previously described (Soriano et al.,transforming growth factor-b receptors. J. Biol. Chem. 270, 21144–1991) and were the generous gift of Drs. P. Soriano and P. Stein.21150.Additional stocks were obtained from Jackson Labs. Mouse ears

were injected intradermally (Eriksson et al., 1980) with 5 ml of adeno- Connolly, D.T., Heuvelman, D.M., Nelson, R., Olander, J.V., Eppley,virus expressing either VEGF or b-galactosidase and the ears photo- B.L., Delfino, J.J., Siegel, N.R., Leimgruber, R.M., and Feder, J.graphed after 5 days with a stereoscope. (1989). Tumor vascular permeability factor stimulates endothelial

The Miles assay (Miles and Miles, 1952) was adapted for mice by cell growth and angiogenesis. J. Clin. Invest. 84, 1470–1478.injecting 10 ml of VEGF (400 ng/ml), allyl isothiocyanate (mustard Courtneidge, S.A., Fumagalli, S., Koegl, M., Superti-Furga, G., andoil, 20% v/v in mineral oil), or saline intradermally into mice that had Twamley-Stein, G.M. (1993). The Src family of protein tyrosine ki-previously been intravenously injected with 100 ml of 0.5% Evan’s nases: regulation and functions. Dev. Suppl., 57–64.blue. After 15 min, the skin patches were dissected, photographed, D’Amore, P.A. (1994). Mechanisms of retinal and choroidal neovas-and eluted at 568C with formalin and quantitated with a spectropho- cularization. Invest. Ophthalmol. Vis. Sci. 35, 3974–3979.tometer (OD600).

Dvorak, H.F., Brown, L.F., Detmar, M., and Dvorak, A.M. (1995).Vascular permeability factor/vascular endothelial growth factor, mi-

Intracerebral Injection and Determinationcrovascular hyperpermeability, and angiogenesis. Am. J. Pathol.

of Blood–Brain Barrier Disruption 146, 1029–1039.Saline or VEGF (200 ng in 2 ml) was injected stereotactically into the

Eliceiri, B.P., Klemke, R., Stromblad, S., and Cheresh, D.A. (1998).left or right frontal lobe 92 mm to the left/right of the midline, 0.5Integrin avb3 requirement for sustained mitogen-activated proteinmm rostral from bregma, and 3 mm in depth from the dura, respec-kinase activity during angiogenesis. J. Cell. Biol. 140, 1255–1263.tively. The animals received an Evan’s blue solution intravenouslyEriksson, E., Boykin, J.V., and Pittman, R.N. (1980). Method for in30 min after injection, as described above. After an additional 30vivo microscopy of the cutaneous microcirculation of the hairlessmin, the mice were perfused and the brains were removed. Evan’smouse ear. Microvasc. Res. 19, 374–379.blue fluorescence was observed using confocal laser microscopyFolkman, J., and Shing, Y. (1992). Angiogenesis. J. Biol. Chem. 267,of fresh unfixed cryosections of the brain.10931–10934.

Acknowledgments Fong, G.H., Rossant, J., Gertsenstein, M., and Breitman, M.L. (1995).Role of the Flt-1 receptor tyrosine kinase in regulating the assemblyof vascular endothelium. Nature 376, 66–70.We thank Tessa Brodhag and Catherine Andrews for expert techni-

cal assistance and Drs. Harold Varmus, Peter K. Vogt, and Bing Ferrara, N., and Davis-Smyth, T. (1997). The biology of vascularJiang for helpful discussions, Dr. K. Spencer (TSRI) for the anti- endothelial growth factor. Endocr. Rev. 18, 4–25.vWF immunofluorescence, R. Xiang and C. Dolman for intravenous Friedlander, M., Brooks, P.C., Shaffer, R.W., Kincaid, C.M., Varner,injections (TSRI), Archana Reddy for the VEGF adenovirus (TSRI), J.A., and Cheresh, D.A. (1995). Definition of two angiogenic path-and Ana Venegas (NIH) for assistance with mouse breeding. We ways by distinct av integrins. Science 270, 1500–1502.also thank Drs. R. Reisfeld (TSRI) and H. Varmus (NIH) for critical Hughes, S.H., Greenhouse, J.J., Petropoulos, C.J., and Sutrave, P.reading of this manuscript. Chick CAM and mouse experiments (1987). Adaptor plasmids simplify the insertion of foreign DNA intowere conducted in accordance with institutional and NIH guidelines. helper-independent retroviral vectors. J. Virol. 61, 3004–3012.B. P. E. was supported by an NIH NRSA postdoctoral fellowship

Inoue, H., Asaka, T., Nagata, N., and Koshihara, Y. (1997). Mecha-(1F32HL09435), R. P. by Deutsche Forschungs Gemeinschaft Panism of mustard oil–induced skin inflammation in mice. Eur. J. Phar-

749/1–1, J. D. H. by an NIH training grant (1T32CA75924), J. L. bymacol. 333, 231–240.

an Army Breast Cancer Program (DAMD179616104), and D. A. C.Kaplan, K.B., Bibbins, K.B., Swedlow, J.R., Arnaud, M., Morgan,by grants CA50286, CA45726, HL54444, and P01 CA78045 from theD.O., and Varmus, H.E. (1994). Association of the amino-terminalNIH. This is manuscript 11851-IMM from The Scripps Researchhalf of c-Src with focal adhesions alters their properties and isInstitute.regulated by phosphorylation of tyrosine 527. EMBO J. 13, 4745–4756.

Received June 30, 1999; revised September 30, 1999.Kiefer, F., Anhauser, I., Soriano, P., Aguzzi, A., Courtneidge, S.A., andWagner, E.F. (1994). Endothelial cell transformation by polyomavirus

References middle T antigen in mice lacking Src-related kinases. Curr. Biol. 4,100–109.

Bett, A.J., Haddara, W., Prevec, L., and Graham, F.L. (1994). AnKim, K.J., Li, B., Winer, J., Armanini, M., Gillett, N., Phillips, H.S., andefficient and flexible system for construction of adenovirus vectorsFerrara, N. (1993). Inhibition of vascular endothelial growth factor-with insertions or deletions in early regions 1 and 3. Proc. Natl.induced angiogenesis suppresses tumour growth in vivo. Nature

Acad. Sci. USA 91, 8802–8806.362, 841–844.

Brooks, P.C., Clark, R.A., and Cheresh, D.A. (1994a). RequirementKlinghoffer, R.A., Sachsenmaier, C., Cooper, J.A., and Soriano, P.

of vascular integrin aVb3 for angiogenesis. Science 264, 569–571. (1999). Src family kinases are required for integrin but not PDGFRBrooks, P.C., Montgomery, A.M., Rosenfeld, M., Reisfeld, R.A., Hu, signal transduction. EMBO J. 18, 2459–2471.T., Klier, G., and Cheresh, D.A. (1994b). Integrin aVb3 antagonists Meredith, J.E., Jr., Fazeli, B., and Schwartz, M.A. (1993). The extra-promote tumor regression by inducing apoptosis of angiogenic cellular matrix as a cell survival factor. Mol. Biol. Cell 4, 953–961.blood vessels. Cell 79, 1157–1164.

Miles, A.A., and Miles, E.M. (1952). Vascular reactions to histamine,Broome, M.A., and Hunter, T. (1996). Requirement for c-Src catalytic histamine-liberator and leukotaxine in the skin of guinea pigs. J.activity and the SH3 domain in platelet-derived growth factor BB Physiol. 118, 228–257.and epidermal growth factor mitogenic signaling. J. Biol. Chem. 271, Miller, J.W., Adamis, A.P., Shima, D.T., D’Amore, P.A., Moulton, R.S.,16798–16806. O’Reilly, M.S., Folkman, J., Dvorak, H.F., Brown, L.F., Berse, B., etBull, H.A., Brickell, P.M., and Dowd, P.M. (1994). Src-related protein al. (1994). Vascular endothelial growth factor/vascular permeabilitytyrosine kinases are physically associated with the surface antigen factor is temporally and spatially correlated with ocular angiogen-CD36 in human dermal microvascular endothelial cells. FEBS Lett. esis in a primate model. Am. J. Pathol. 145, 574–584.351, 41–44. Murohara, T., Horowitz, J.R., Silver, M., Tsurumi, Y., Chen, D., Sulli-

van, A., and Isner, J.M. (1998). Vascular endothelial growth factor/Chang, M.W., Barr, E., Lu, M.M., Barton, K., and Leiden, J.M. (1995).

Molecular Cell924

vascular permeability factor enhances vascular permeability via ni-tric oxide and prostacyclin. Circulation 97, 99–107.

Nada, S., Okada, M., MacAuley, A., Cooper, J.A., and Nakagawa,H. (1991). Cloning of a complementary DNA for a protein-tyrosinekinase that specifically phosphorylates a negative regulatory site ofp60c-src. Nature 351, 69–72.

Passaniti, A., Taylor, R.M., Pili, R., Guo, Y., Long, P.V., Haney, J.A.,Pauly, R.R., Grant, D.S., and Martin, G.R. (1992). A simple, quantita-tive method for assessing angiogenesis and antiangiogenic agentsusing reconstituted basement membrane, heparin, and fibroblastgrowth factor. Lab. Invest. 67, 519–528.

Satake, S.M., Kuzuya, M., et al. (1998). Angiogenic stimuli are essen-tial for survival of vascular endothelial cells in three-dimensionalcollagen lattice. Biochem. Biophys. Res. Comm. 244, 642–646.

Schwartz, M.A., Schaller, M.D., and Ginsberg, M.H. (1995). Integrins:emerging paradigms of signal transduction. Annu. Rev. Cell. Dev.Biol. 11, 549–599.

Senger, D.R., Galli, S.J., Dvorak, A.M., Perruzzi, C.A., Harvey, V.S.,and Dvorak, H.F. (1983). Tumor cells secrete a vascular permeabilityfactor that promotes accumulation of ascites fluid. Science 219,983–985.

Shalaby, F., Rossant, J., Yamaguchi, T.P., Gertsenstein, M., Wu,X.F., Breitman, M.L., and Schuh, A.C. (1995). Failure of blood-islandformation and vasculogenesis in Flk-1-deficient mice. Nature 376,62–66.

Soriano, P., Montgomery, C., Geske, R., and Bradley, A. (1991).Targeted disruption of the c-src proto-oncogene leads to os-teopetrosis in mice. Cell 64, 693–702.

Stein, P.L., Vogel, H., and Soriano, P. (1994). Combined deficienciesof Src, Fyn, and Yes tyrosine kinases in mutant mice. Genes Dev.8, 1999–2007.

Stoker, A.W., Hatier, C., and Bissell, M.J. (1990). The embryonicenvironment strongly attenuates v-src oncogenesis in mesenchymaland epithelial tissues, but not in endothelia. J. Cell Biol. 111,217–228.

Takenaka, H., Kishimoto, S., Tooyama, I., Kimura, H., and Yasuno,H. (1997). Protein expression of fibroblast growth factor receptor–1in keratinocytes during wound healing in rat skin. J. Invest. Dermatol.109, 108–112.

Thomas, S.M., and Brugge, J.S. (1997). Cellular functions regulatedby Src family kinases. Annu. Rev. Cell. Dev. Biol. 13, 513–609.

Twamley-Stein, G.M., Pepperkok, R., Ansorge, W., and Courtneidge,S.A. (1993). The Src family tyrosine kinases are required for platelet-derived growth factor–mediated signal transduction in NIH 3T3 cells.Proc. Natl. Acad. Sci. USA 90, 7696–7700.

Weidner, N., Semple, J.P., Welch, W.R., and Folkman, J. (1991).Tumor angiogenesis and metastasis—correlation in invasive breastcarcinoma. N. Engl. J. Med. 324, 1–8.

Ziche, M., Morbidelli, L., Choudhuri, R., Zhang, H.T., Donnini, S.,Granger, H.J., and Bicknell, R. (1997). Nitric oxide synthase liesdownstream from vascular endothelial growth factor–induced butnot basic fibroblast growth factor–induced angiogenesis. J. Clin.Invest. 99, 2625–2634.